mcpField
Q
qualys-mcp

qualys-mcp

Local
github.com/rijul170Security0Updated 15w ago

About qualys-mcp

qualys-mcp is an MCP server for managing Qualys security products and administration through MCP clients. It spans multiple Qualys API families and gates destructive operations behind explicit opt-in and confirmation.

Key features

  1. 01Coverage across Qualys product modules
  2. 02Tools mapped to documented API operations
  3. 03FO, QPS REST, and Gateway APIs
  4. 04Read, write, and destructive safety tiers
  5. 05Destructive tools require opt-in and confirmation
  6. 06Credentials supplied through environment variables or Windows DPAPI

Use cases

  1. 01Review VMDR findings and host detections
  2. 02Launch scans and search WAS findings
  3. 03Manage Cloud Agent and container security data
  4. 04Limit a console to licensed modules

MCP support

Exposes Qualys product and administration operations as tools for MCP clients. Each server process is configured for a Qualys console.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"rijul170-qualys": {
"command": "uvx",
"args": [
"qualys-mcp-full"
]
}
}
}
Run commanduvx qualys-mcp-full

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
pypiqualys-mcp-full0.1.0

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

Qualys MCP

A Model Context Protocol server for full Qualys portal management — expose VMDR, Policy Compliance, WAS, Cloud Agent, Container Security, TotalCloud, Patch Management, CSAM/GAV, EASM and administration to any MCP‑capable client (Claude, and other MCP hosts).

CI Python License: MIT MCP Tools

One process per Qualys console (run as many as you have subscriptions). Built on FastMCP, with a single client that speaks all three Qualys API families, a strict read / write / destructive safety model, and credentials that never touch chat or plaintext.


Highlights

  • Broad coverage — 59 auto‑discovered feature modules spanning the whole portal: VMDR, PC, WAS, Cloud Agent, Container Security, TotalCloud/CloudView, Patch Management, CSAM/GAV, EASM, FIM, EDR, PCRS, and administration.
  • ~400 tools — list_hosts, list_host_detections, launch_scan, search_was_findings, list_knowledgebase, get_cs_image_vulnerabilities, and hundreds more, each mapped 1:1 to a documented Qualys API operation.
  • Three API regimes, one client — the classic "FO" XML API, the QPS REST API, and the JWT‑authenticated Gateway are all handled behind self._fo / self._qps / self._gateway.
  • Safety tiers are load‑bearing — every tool is annotated read‑only, write, or destructive. Destructive tools (delete / purge / deactivate) are not even registered unless you opt in per console, and still require a per‑call confirmation token.
  • Lazy auth — no network calls at import or construction; --check runs fully offline with no credentials.
  • Secrets stay secret — on Windows, credentials live only in DPAPI‑encrypted blobs; on any OS you can supply them via environment variables. Nothing is ever written to the repo.

The three API regimes

RegimeBase pathAuthPayload
Classic "FO"/api/2.0/fo/, /msp/Basic + X-Requested-With headerform → XML
QPS REST/qps/rest/BasicXML <ServiceRequest> / JSON
Gatewaygateway.<pod>.apps.qualys.comBearer JWT (from /auth)JSON

Architecture

qualys_mcp/
  server.py     FastMCP server; lazy auth; --check offline validation
  registry.py   auto-discovers modules/*.py
  client.py     one client, three API regimes (FO / QPS / Gateway-JWT)
  config.py     per-console config + platform (POD) -> URL map
  common/       auth (HTTP mw), errors, logging, rate_limit, utils, xml
  modules/      base.py + one file per feature module (the bulk of the surface)

Adding a capability is just dropping a file in qualys_mcp/modules/ — see docs/MODULE_BUILD_GUIDE.md.

Quick start

1. Install
python -m venv .venv
# Windows:      .\.venv\Scripts\pip install -e .
# macOS/Linux:  ./.venv/bin/pip install -e .
2. Offline sanity check (no credentials required)
python -m qualys_mcp --check

Lists every discovered module with its tool count and exits — a good first smoke test.

3. Provide credentials

You need a Qualys API user and your platform/POD code (e.g. US1, US2, EU1; find it under your console URL or Help → About).

Any OS — environment variables:

export QUALYS_USERNAME='api-user'
export QUALYS_PASSWORD='api-pass'
export QUALYS_PLATFORM='US2'          # or set QUALYS_API_URL / QUALYS_GATEWAY_URL
python -m qualys_mcp --transport streamable-http --host 127.0.0.1 --port 8781 --console-label consulting

Windows — DPAPI‑encrypted blobs (recommended for a persistent deployment):

# Encrypt once per console; the blob is user-scoped and stored under .secrets/
.\encrypt-qualys-creds.ps1 -Console consulting -Platform US2
.\encrypt-qualys-creds.ps1 -Console cloud      -Platform US2

# Launch one instance per console (:8781, :8782)
.\start-qualys-mcp.ps1
#   ...opt into destructive tools for a console only when you mean it:
.\start-qualys-mcp.ps1 -EnableDestructive cloud

# Optional: keep both alive across reboots via a Scheduled Task watchdog
.\register-qualys-mcp-task.ps1

See .env.example for every QUALYS_* setting.

4. Point your MCP client at it

A ready‑to‑use .mcp.json is included for Claude Code:

{
  "mcpServers": {
    "qualys-consulting": { "type": "http", "url": "http://localhost:8781/mcp" },
    "qualys-cloud":      { "type": "http", "url": "http://localhost:8782/mcp" }
  }
}

scripts/register_mcp_clients.py shows how to register the same local‑HTTP servers into several MCP clients at once.

Safety model

  • Tools are annotated read‑only, write, or destructive.
  • Destructive tools (delete, purge, deactivate, uninstall) are not registered unless the console runs with QUALYS_ENABLE_DESTRUCTIVE=true.
  • Even when enabled, each destructive tool requires a matching confirm=<id> argument before it will act — so a destructive call can never happen by accident from a single prompt.

Read docs/SAFETY_AUDIT.md for the full tiering.

Module & tool inventory

  • 59 modules auto‑discovered from qualys_mcp/modules/.
  • ~360 tools registered by default; ~400 with QUALYS_ENABLE_DESTRUCTIVE=true (the extra ~40 are the guarded destructive tools that otherwise stay hidden).
  • Spread across the three API regimes, with a handful of modules spanning two.

The full per‑module table (registry name, API family, tool counts, one‑line description) is in docs/MODULE_INDEX.md, and a complete endpoint → tool → action breakdown is in docs/ENDPOINTS_TOOLS_ACTIONS.md.

Configuration reference

VariablePurpose
QUALYS_USERNAME / QUALYS_PASSWORDAPI user credentials
QUALYS_PLATFORMPOD code (US1…KSA1) — auto‑fills API + Gateway URLs
QUALYS_API_URL / QUALYS_GATEWAY_URLExplicit overrides for unlisted PODs
QUALYS_CONSOLE_LABELNames the console in logs and the server name
QUALYS_ENABLE_DESTRUCTIVEtrue to register the destructive tier
QUALYS_MCP_MODULESComma‑separated allowlist to scope a console to its licensed modules
QUALYS_MCP_API_KEYOptional shared secret for the HTTP transport

Development

pip install -e ".[dev]"
python -m qualys_mcp --check   # offline validation
pytest                          # unit tests (no network)
ruff check .                    # lint

Contributions welcome — see CONTRIBUTING.md.

Disclaimer

This is an independent, community project. It is not affiliated with, endorsed by, or supported by Qualys, Inc. "Qualys" and product names are trademarks of their respective owners. Use against your own subscriptions in accordance with your Qualys license and API terms. The software is provided "as is" (see LICENSE); you are responsible for what you run against your environment — especially the destructive tier.

Star history

Star History Chart

License

MIT © 2026 Rijul Sharma

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.